As far as speed, R is a lot like similar numeric tools/languages. It's designed to be very high level, building on top of heavily optimized subroutines written in lower level languages. This is why the kind of comparisons in the article are silly: almost every language will end up being the same speed because they're all just calling the same basic BLAS routines N times with virtually zero overhead (relative to the inner loop).
This book helped me a lot because it goes into how the core language works, something that is (sadly) almost always ignored by people who learn and teach R: http://www.amazon.com/Software-Data-Analysis-Programming-Sta...
As far as speed, R is a lot like similar numeric tools/languages. It's designed to be very high level, building on top of heavily optimized subroutines written in lower level languages. This is why the kind of comparisons in the article are silly: almost every language will end up being the same speed because they're all just calling the same basic BLAS routines N times with virtually zero overhead (relative to the inner loop).